Back to results
US 11,537,858 B2Granted
Computing device and method
AssigneeShanghai Cambricon Information Technology Co., Ltd.
InventorsTianshi Chen, Xuda Zhou, Shaoli Liu, Zidong Du
Life of this patent
Estimating the term…
Abstract
A computing device, comprising: a computing module, comprising one or more computing units; and a control module, comprising a computing control unit, and used for controlling shutdown of the computing unit of the computing module according to a determining condition. Also provided is a computing method. The computing device and method have the advantages of low power consumption and high flexibility, and can be combined with the upgrading mode of software, thereby further increasing the computing speed, reducing the computing amount, and reducing the computing power consumption of an accelerator.
Claims15
A processor, comprising:
- an instruction control unit configured to fetch a processing instruction;
- an operation module configured to receive frame information, neural network parameters, and the processing instruction, and perform neural network operations on the frame information and the neural network parameters according to the processing instructions;
- and
- a storage module configured to store the frame information and the neural network parameters,
- wherein the frame information includes complete frame information and reference frame information, and the neural network parameters include neurons, weights, topological structures and/or processing instructions, and
- wherein the operation module includes:
- an accurate operation unit configured to:
- fetch the complete frame information and weights in the neural network parameters,
- perform neural network operations to obtain a first operation result, and
- transfer the first operation result to the storage module, and
- an approximate operation unit configured to:
- fetch the reference frame information and an operation result of the reference frame which is obtained in advanced and stored in the storage module,
- perform approximate operations to obtain a second operation result, and transfer the second operation result to the storage module.
10 more elements
A processing method comprising:
- fetching a processing instruction, by an instruction control unit, to control an operation module;
- and
- receiving, by the operation module, frame information, neural network parameters and the processing instruction, and performing neural network operations on the frame information and the neural network parameters according to the processing instruction;
- wherein the frame information and the neural network parameters are stored in a storage module,
- wherein the frame information includes complete frame information and reference frame information, and the neural network parameters include neurons, weights, topological structures and/or processing instructions, and
- wherein the operation module includes an accurate operation unit and an approximate operation unit, and the receiving, by the operation module, the frame information, the neural network parameters and the processing instruction, and performing neural network operations on the frame information and the neural network parameters according to the processing instruction includes:
- fetching, by the accurate operation unit, the complete frame information and weights in the neural network parameters, performing neural network operations to obtain a first operation result, and transferring the first operation result to the storage module, and
- fetching, by the approximate operation unit, the reference frame information and an operation result of the reference frame which is obtained in advanced and stored in the storage module, performing approximate operations to obtain a second operation result, and transferring the second operation result to the storage module.
5 more elements
An electronic device, comprising:
- a processor that includes:
- an instruction control unit configured to fetch a processing instruction;
- an operation module configured to receive frame information, neural network parameters, and the processing instruction, and perform neural network operations on the frame information and the neural network parameters according to the processing instructions;
- and
- a storage module configured to store the frame information and the neural network parameters,
- wherein the frame information includes complete frame information and reference frame information, and the neural network parameters include neurons, weights, topological structures and/or processing instructions, and
- wherein the operation module includes:
- an accurate operation unit configured to:
- fetch the complete frame information and weights in the neural network parameters,
- perform neural network operations to obtain a first operation result, and
- transfer the first operation result to the storage module, and
- an approximate operation unit configured to:
- fetch the reference frame information and an operation result of the reference frame which is obtained in advanced and stored in the storage module,
- perform approximate operations to obtain a second operation result, and transfer the second operation result to the storage module.
11 more elements
Dependencies are inferred from the claim text.